Morphowiki

/new-word — 語素派生與自動造詞

隨著您收錄的單字增加，字典中已累積豐富的字首、字根與字尾資源。本指令會自動掃描語素池，進行排列組合，挖掘出您尚未收錄但真實存在的英文單字，並為您一鍵完成建檔。

/new-word — Auto-generate words from existing morphemes

After a few /ingest runs, your dictionary already has a pool of prefixes, roots, and suffixes. This command recombines them to surface real English words you haven't yet collected, and writes the dictionary pages for you in one go.

功能核心：強化語素連結

死背課本單字表僅能擴張零散的記憶點；從已習得的語素出發，延伸出整族同根詞，才是真正內化「語素分析」思維的關鍵。

假設您已學過 believe、-able、un-、re-，系統會輔助引導您將 believable、unbelievable、believer、disbelieve 等派生詞納入學習版圖。手動執行 /ingest 雖然可行，但 /new-word 能更主動地為您探索潛在的學習目標。

/new-word 將您字典中現有的 morpheme（語素）視為樂高積木，自動拼湊出真實單字。您只需決定哪些單字具備複習價值，即可輕鬆擴充您的字根字首知識庫。

Why this exists

Drilling a textbook vocab list is one thing; surfacing the productive derivations of words you've already learned is what cements morpheme literacy.

Say you have already ingested believe, -able, un-, re-. The natural next move is believable, unbelievable, believer, disbelieve. You can keep typing /ingest one word at a time, but you also need to remember "right, there's also disbelieve."

/new-word remembers for you: it treats your dictionary's existing morphemes (the smallest meaningful units of a word) as Lego bricks, recombines them into real English words, and lets you pick which ones are worth reviewing.

具體範例

假設你之前 ingest 過 unbelievable、predictable、international,這時字典的 morpheme 池大概像這樣:

  • 字首:un-、re-、inter-
  • 字根:believe、predict、nat
  • 字尾:-able、-ion、-al

在 Claude Code 對話框打:

/new-word 3

幾秒後,Claude 回報它生成的三個字(實際結果會因池內容與當次取樣而不同,以下是可能的一種):

要求生成: 3
實際生成: 3 個

- believable — 可信的 [root=believe, suffix=-able]
- prediction — 預測;預言 [root=predict, suffix=-ion]
- national   — 國家的;全國性的 [root=nat, suffix=-ion, suffix=-al]

下一步建議:/flashcard believable 把這些字加進複習庫。

同時,以下事情自動發生:

  • 三個新檔 dictionary/words/believable.md、prediction.md、national.md 建好,內容格式與 /ingest 完全一樣。
  • 對應 morpheme 檔(root/believe.md、suffix/-able.md、root/predict.md、suffix/-ion.md、root/nat.md、suffix/-al.md)的 ## Words containing this 區段各自加入反向連結。
  • dictionary/index.md 與 dictionary/log.md 同步更新(log 動詞會是 new-word 而不是 ingest,方便日後追溯來源)。

新建的 dictionary/words/believable.md 內容(節錄)長這樣:

---
word: believable
pos: adj.
ipa: /bɪˈliː.və.bəl/
added: 2026-05-24
---

# believable

**詞性**: adj.
**IPA**: /bɪˈliː.və.bəl/
**中文釋義**: 可信的;令人信服的

## 拆解
- 字根: [[../root/believe]] `believe` — 相信
- 字尾: [[../suffix/-able]] `-able` — 形容詞字尾,「可…的」

## 詞源
…(2–4 句說明 believe 自古英語 belyfan 而來等)

## 記憶法
…(1–3 句連結語素或意象的記憶提示)

## 筆記區

關於 section 名稱:檔案內的 ## 拆解 / 詞源 / 記憶法 / 筆記區 永遠是中文,因為本機複習網頁是按這幾個字串去解析內容的。field 標籤 **詞性** / **IPA** / **中文釋義** 也固定中文。不要改名;改了之後 /review-word 會找不到內容。釋義文字本身會隨 CLAUDE.md 的語系設定改變。

A concrete example

Say you've previously ingested unbelievable, predictable, and international. Your dictionary's morpheme pool now looks roughly like this:

  • Prefixes: un-, re-, inter-
  • Roots: believe, predict, nat
  • Suffixes: -able, -ion, -al

In the Claude Code chat, type:

/new-word 3

After a few seconds, Claude reports the three words it generated (actual output varies with pool contents and sampling — this is one possible run):

Requested:  3
Generated:  3

- believable — possible to believe [root=believe, suffix=-able]
- prediction — a forecast or statement about the future [root=predict, suffix=-ion]
- national   — relating to a nation [root=nat, suffix=-ion, suffix=-al]

Next: /flashcard believable to add these to your review deck.

At the same time, these things happen automatically:

  • Three new files dictionary/words/believable.md, prediction.md, national.md are created, in the same format as /ingest.
  • The matching morpheme files (root/believe.md, suffix/-able.md, root/predict.md, suffix/-ion.md, root/nat.md, suffix/-al.md) each get a back-link added under ## Words containing this.
  • dictionary/index.md and dictionary/log.md are updated (the log verb is new-word rather than ingest, so you can trace the source later).

The new dictionary/words/believable.md looks roughly like this (abridged):

---
word: believable
pos: adj.
ipa: /bɪˈliː.və.bəl/
added: 2026-05-24
---

# believable

**詞性**: adj.
**IPA**: /bɪˈliː.və.bəl/
**中文釋義**: 可信的;令人信服的

## 拆解
- 字根: [[../root/believe]] `believe` — to trust as true
- 字尾: [[../suffix/-able]] `-able` — adjective suffix, "capable of"

## 詞源
…(two to four sentences on the history of believe, from Old English belyfan, etc.)

## 記憶法
…(one to three sentences of a mnemonic linking the morphemes)

## 筆記區

About section names: the headings ## 拆解 / 詞源 / 記憶法 / 筆記區 in the file stay in Chinese because the review web app parses them by name. The field labels **詞性** / **IPA** / **中文釋義** are also fixed Chinese. Don't rename them — if you do, /review-word won't find the content. The prose values inside follow the locale set in CLAUDE.md.

操作流程

  1. 累積基礎語素：理想情況下，建議先執行 5–10 次 /ingest，確保語素資料夾（prefix、root、suffix）中已累積一定的基礎內容。
  2. 啟動指令：在終端機輸入 /new-word 開始生成。 
    /new-word        # 預設生成 1 個單字
/new-word 5      # 生成 5 個單字
/new-word 10     # 一次生成 10 個單字（上限 20）
  3. 審閱結果：查看系統回報的單字清單、拆解邏輯與簡短釋義。
  4. 檢查檔案：逐一打開 dictionary/words/<word>.md 檢查內容。若對某單字不滿意，直接刪除該檔案即可。
  5. 納入複習：確認無誤後，使用 /flashcard <word> 將單字加入複習庫。

實際生成效能：若請求數量超過語素池所能組合的上限，系統將如實回報實際生成的數量。這通常意味著您需要透過 /ingest 增加更多新的基礎語素。

Step by step

  1. Make sure your dictionary already has some morphemes. Ideally run /ingest 5–10 times first, so that dictionary/prefix/, dictionary/root/, dictionary/suffix/ each contain a handful of files.
  2. Open Claude Code from inside the project folder (run claude in your terminal).
  3. In the chat box, type one of: 
    /new-word        # default: 1 word
/new-word 5      # 5 words
/new-word 10     # 10 at a time (max 20)
  4. Wait for Claude's report. It tells you how many words it actually produced, with each word's breakdown and a short definition.
  5. Open dictionary/words/<new-word>.md for each and check the content. Don't like one? Just delete the file — next time you run /lint it will flag any stray morpheme back-links to clean up.
  6. For the ones you like, run /flashcard <word> to add them to your review deck.

How many to ask for: early on, when your morpheme pool is small, /new-word 3 is plenty. Once the pool grows, run /new-word 10 and cherry-pick from the report.

You may get fewer than requested: if you ask for /new-word 10 but the pool only allows 6 valid combinations, Claude reports "requested 10, generated 6" honestly and explains the pool ran out. Go run more /ingest calls to grow the morpheme pool, then try again.

常見問題與限制

非隨機造詞工具：/new-word 僅會生成真實存在的英文單字。若您的語素池無法組出有效的詞彙，系統將回報生成數量為 0。它不會強行拼湊不存在的單字。

與 /ingest 的區別：/ingest 是「精確建檔」，由您指定單字；/new-word 是「潛力挖掘」，由系統根據現有基礎主動推薦單字。

單層建檔邏輯：/new-word 僅會為生成的 N 個單字建檔，不會像 /ingest 一樣遞迴拆解中間單字。這賦予您對字典內容更精確的控制權。

語素池門檻：若您的資料夾中尚無語素（Prefix/Root/Suffix），指令將無法執行。請先收錄一些單字以建立初始語素庫。

下一步：

← 回首頁

Common mistakes & next steps

Mistake 1 — Expecting Claude to invent new words: /new-word will not generate fake combinations like sushify or unsalty. Its strict rule is "every word produced must be an attested English word." If it can't compose one, it honestly reports zero. This is not the right tool for coining neologisms.

Mistake 2 — Confusing it with /ingest: /ingest <word> means "you pick a word and I'll add it"; /new-word means "you don't pick — I mine the existing morpheme pool for words you haven't collected yet." The first hands you a fish; the second teaches you what other fish are in your pond.

Mistake 3 — Assuming it recurses into intermediate words like /ingest does: /ingest international also builds dictionary pages for national and nation; but /new-word only writes the N words it explicitly generated — it does not auto-expand into shorter forms inside them. This is on purpose: you chose precise control over count. If you want the intermediates, follow up with /ingest.

Mistake 4 — Running it on an empty pool: if any of the three pools is empty (no files at all), /new-word will refuse and tell you to ingest a few words first. With only 2–3 morphemes in the pool, valid combinations run out fast and you may only get 1–2 words. Don't be discouraged — ingest a few more and try again.

Next steps:

  • Add the generated words to your review deck: /flashcard.
  • Verify all two-way links are still consistent: /lint.
  • To remove a word you don't like, or swap in a better one, use /ingest for the replacement and delete the unwanted .md by hand in your file manager.
  • Start reviewing: /review-word.

Back to home.

Common mistakes & next steps

Mistake 1 — Expecting Claude to invent new words: /new-word will not generate fake combinations like sushify or unsalty. Its strict rule is "every word produced must be an attested English word." If it can't compose one, it honestly reports zero. This is not the right tool for coining neologisms.

Mistake 2 — Confusing it with /ingest: /ingest <word> means "you pick a word and I'll add it"; /new-word means "you don't pick — I mine the existing morpheme pool for words you haven't collected yet." The first hands you a fish; the second teaches you what other fish are in your pond.

Mistake 3 — Assuming it recurses into intermediate words like /ingest does: /ingest international also builds dictionary pages for national and nation; but /new-word only writes the N words it explicitly generated — it does not auto-expand into shorter forms inside them. This is on purpose: you chose precise control over count. If you want the intermediates, follow up with /ingest.

Mistake 4 — Running it on an empty pool: if any of the three pools is empty (no files at all), /new-word will refuse and tell you to ingest a few words first. With only 2–3 morphemes in the pool, valid combinations run out fast and you may only get 1–2 words. Don't be discouraged — ingest a few more and try again.

Next steps:

Back to home.